Luckily I haven't ran into replication much. Because as the only technology seems to be the defacto where I work is clustering or nothing heh. But for replication issues; I would think Identity column causes issues in Merge replications correct? Because seed values might be different in two different nodes?
But aside from that in stand alone databases, mirroring, etc. Is there anywhere you guys have found Identity column to be a bad choice? Thanks.
Merge is just one of the problems, in transactional you will have to checkident the subscriber tables if you need failover.
Disconnected Editing is cumbersome with identity columns; therefore ID tables are there to rescue you
.. and there are more but they are very specific and you could arguably say that identity could still be used.
Don't get me wrong on places like DW with Dimession and Fact Tables it makes sense.
On OLTP and disconnected systems, there are many situations in which there are better alternatives.